Agent skills for Glyphs App

Glyphs MCP gives an AI client safe, structured access to the Glyphs canvas. Agent skills add the missing layer on top: reusable workflows for real font-editing tasks.

Think of it this way:

• MCP tools expose what the plug-in can do
• skills teach an agent how to use those tools well for a specific job

That matters in Glyphs because font editing is stateful. A good workflow should inspect the current font and master first, prefer dedicated tools over free-form code, use dry runs before mutations, and avoid auto-saving unless a human explicitly asks for it.

Why skills help in Glyphs

Glyphs workflows often have a few sharp edges:

• the same prompt can target the wrong font or master if the agent does not read context first
• spacing and kerning changes should usually be reviewed before they are applied
• outline edits should prefer dedicated tools and bounded docs lookups over broad code execution
• large tool lists can waste tokens in MCP clients unless you pick the right profile first

The bundled Glyphs MCP skills encode those habits so agents reach for safer defaults without repeating the whole procedure every session.

Available skills

This repo currently ships a small curated set of skills for real Glyphs MCP tasks:

Skill	Use it for	Example prompt
glyphs-mcp-connect	Start-up checks, MCP connection, and first health checks	Use the glyphs-mcp-connect skill to verify my setup and call list_open_fonts.
glyphs-mcp-kerning	Kerning bumper review and approved applies	Use the glyphs-mcp-kerning skill to review bumper suggestions and do a dry run first.
glyphs-mcp-spacing	Spacing review, sidebearing suggestions, and approved applies	Use the glyphs-mcp-spacing skill to review spacing for the selected glyphs and do a dry run first.
glyphs-mcp-outlines-docs	Outline edits, components, anchors, selected nodes, and docs lookup	Use the glyphs-mcp-outlines-docs skill to inspect selected nodes, prefer dedicated tools, and fetch only the docs I need.

These are intentionally narrow. The goal is not to create a single mega-skill for all of Glyphs. The goal is to make the common workflows predictable and safe.

How the bundle is exposed

The authored source of truth lives in skills/.

For contributors and advanced users, the repo exposes that same skill bundle to both clients through lightweight bridges:

• Codex: .agents/skills -> ../skills
• Claude Code: .claude/skills -> ../skills

That keeps the skill content in one place while letting each client discover it through the path it expects.

For end users, the Glyphs MCP installers can also copy the bundled glyphs-mcp-* skills into the clients’ global skill directories:

• Codex: ~/.codex/skills/
• Claude Code: ~/.claude/skills/

That means you do not need to open the repo just to use the bundled Glyphs MCP skills.

Safety-first workflow

The current skill set follows the same high-level operating model:

1. Pick the narrowest useful Glyphs MCP tool profile first.
2. Read current state before changing anything.
3. Prefer dedicated tools over free-form code.
4. Use review or dry-run steps before apply steps.
5. Never auto-save the font.
6. Prefer docs_search plus docs_get instead of broad docs loading.

This keeps the workflows aligned with the Glyphs MCP safety model and with the tool contracts documented elsewhere in the repo.

Next step

For setup and client-specific usage instructions, go to Use skills in Codex and Claude Code.

For the repo source, see the skills directory on GitHub.